1. Field of the Invention
The present invention relates to an iris recognition method, particularly to an iris recognition method, which utilizes a matching pursuit algorithm to simplify the extraction and reconstruction of iris features to reduce the memory space required by each piece of iris feature vector without the penalty of reducing recognition accuracy.
2. Description of the Related Art
Traditionally, people use the token-based mechanism or the knowledge-based mechanism to perform personal identification. In the token-based mechanism, the user uses a physical identification tool he owns to verify his identity, such as a stamp, an ID card, a passport, a credit card or a key. In the knowledge-based mechanism, the user uses the knowledge he owns to verify his identity, such as a password, the ID number, a telephone number or his birthday. As the two traditional identification mechanisms do not use the intrinsic features, the users are apt to meet some inconveniences; for example, the identification tool is lost, stolen, or imitated; the password is forgotten, incorrectly remembered, or deciphered. When the abovementioned cases happen, the traditional identification mechanisms cannot determine whether the person who owns the identification tool or the password is the authentic user. Thus, the traditional identification mechanisms cannot always guarantee the security function. In contrast with the abovementioned traditional identification mechanisms, the biometric technology has a much superior reliability and becomes the developing tendency in the field concerned. The biometric technology is defined as “automated methods of verifying or recognizing the identity of the user based on his intrinsic features, such as his eyeball, face, fingerprint and voice”.
A good biometric identification mechanism should have the following characteristics: (1) generality, i.e. the adopted biometric feature generally exists in all persons; (2) uniqueness, i.e. two persons cannot possess all the same adopted biometric feature; (3) invariability, i.e. the adopted biometric feature is invariable eternally; and (4) measurability, i.e. the adopted biometric feature can be measured and quantified. The abovementioned conditions should be considered in adopting an appropriate biometric feature. Further, in designing a practical biometric identification mechanism, the following factors also should be considered: (1) performance, i.e. the recognition speed and reliability; (2) acceptability, i.e. users are willing or glad to accept the mechanism; and (3) inescapability, i.e. the mechanism is unlikely to be fooled or cheated, and all the non-authenticated persons cannot escape from the detection of the mechanism. Based on the abovementioned conditions and factors, several biometric identification systems have been extensively used or under trial and estimation, including: face shape, face thermal sensation, fingerprint, palmprint, retina, iris and voice identification systems.
So far, the iris identification system has the highest discrimination rate among the biometric identification systems. Further, the complexity and diverseness of the iris makes the iris identification systems very hard to break through. The iris is an annular region between the sclera and the pupil. The iris texture is complicated enough to be used as a biometric signature. For ordinary people, their irises have completely developed at about two years old, and their iris patterns are also fixed then. In the iris identification technology, the iris image is scanned and transformed into digital codes, and an algorithm is used to extract the discriminable feature vectors from the digital codes; then, the feature vectors are stored in a database. In the personal identification, the identity is verified or recognized via comparing the iris features of the testee with the features stored in the database.
In the conventional technology, massive iris feature data are stored in the database to improve the accuracy of iris identification, and it indeed puts a heavy burden on the system. Accordingly, the present invention proposes an iris recognition method, which utilizes a matching pursuit algorithm to simplify the extraction and reconstruction of iris features and reduce the memory space required by each piece of iris feature data without the penalty of recognition accuracy.